Data analysis - similarities and differences between C-end and b-end data analysis

Data analysis is of great value , However, the value of data analysis is different because of the service objects of data analysis . And the service object of data analysis also has a profound impact on our data analysis strategy .

In general , Data service objects are mainly divided into two categories ∶C End users and B End customers .

C End

C End user mainly refers to the vast majority of applications related to life , Or you can browse 、 Access to many web applications .C The user volume of end users is often very large , for example ,2019 year 3 The active users of wechat are close to 10 Billion ," Are you hungry " There are more than ten million active users , Even a very unknown e-commerce website , It also has hundreds of thousands of monthly active users . Because most C End products are free ,C End users can easily access the services provided by the data service provider （ Like downloading a App Or scan a QR code ）. Of course , Users can also easily discard a previously used data service , And replace with other data services .

From the perspective of data analysis ,C The behavior of the end user is complex 、 Its behavior is full of great uncertainty . Data analysts know C There are many laws in the behavior of end users , But in the face of C Massive data brought by end users , It often seems a little " worry " And not knowing what to do . thus , Be a data mining engineer , Or the data algorithm engineer is facing these data , In the service of C End user time , Their degrees of freedom will be relatively large . Data algorithm engineers can use very flexible methods , Using very diverse and complex models , Data processing .

Although the product manager or other business representatives can not give guidance to the data algorithm engineer on how to deal with the data , But this does not mean that the output of data algorithm engineers can be arbitrary . Data analysts often need to match product needs , And meet certain business indicators .

B End

The narrow sense B End customers usually refer to enterprise customers or commercial customers , The generalized B In addition to the above-mentioned enterprise customers, the end customers , It also includes government customers （ Sometimes government clients are individually referred to as G End customers ,G namely Gov Abbreviation ） And other organizational customers .

be relative to C End ,B The number of terminals will be small , but B End customers have some social or organizational influence , Also has a wealth of business knowledge . meanwhile B End and C The feedback of the end to the result is different ,C End users are usually through the actions of user groups , Direct feedback with data ;B On the other hand, they communicate through meetings 、 Business negotiation 、 Direct feedback from customers, such as business contracts .

B The data business of end customers does not rely entirely on data , Customers themselves will put forward relevant business needs and business knowledge , For data analysts , When the amount of data is not enough , These are the requirements and demand knowledge from the business side , It can often help data analysts master the laws of data , Help data analysts with data analysis . Of course , These trivial requirements and requirements may also bind the hands and feet of data analysts . meanwhile , Data Algorithm Engineers mine and summarize abnormal rules from data , Customers will doubt the correctness of the results ; When the output of the data algorithm engineer conforms to common sense , Customers will doubt the necessity of data analysis and mining ……· And so on , Make in relation to B In the business cooperation with end customers , Communicating with each other often consumes a lot of energy , The proportion of data analysis and rule mining will be relatively reduced .

B End customers often require that the output of data analysts be highly interpretable . quite a lot B Although the end customers have very professional business knowledge , But rival data science lacks understanding . When a data algorithm engineer makes a model , Customers often require data algorithm engineers to provide highly interpretable principle descriptions . More customers , In addition to strong interpretability , It also requires that the model produced by the data algorithm engineer be controllable , If the customer thinks that some indicators need to be adjusted , Be able to exert control immediately . Therefore, models with poor interpretability such as neural networks are B It is rarely used in the business scenario of end customers , Like support vector machines （Support Vector Machine,SVM） Although this has strong explicability , But let customers understand and accept that this interpretability requires a model that consumes huge educational costs , Living space is also very limited . therefore , And B The final model of successful cooperation between end customers , In general , Its complexity will not be too high , Data work is not very difficult , The model is very interpretable , Some intuitive statistical indicators will appear very frequently .
The various visual outputs are B A solution that is very easy for end customers to accept . The visualization scheme depends on the form of charts , It has the advantages of strong directness and weak explicability , If you add cool design , Better fit B The needs of end customers . Many companies with data-driven businesses , A large part of the project will include a large display screen , Or a web page full of various visualization modules , This is also B Reflection of the unique characteristics of end customer related businesses .

Reference material ：《 General knowledge of data analysis 》